Am. J. Hum. Genet. 56:835-844, 1995 Epidemiological Evaluation of the Use of Genetics to Improve the Predictive Value of Disease Risk Factors Muin J. Khoury' and Diane K. Wagener2 'Birth Defects and Genetic Diseases Branch, Division of Birth Defects and Developmental Disabilities, National Center for Environmental Health, Atlanta; and 20ffice of Analysis, Epidemiology, and Health Promotion, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD Summary Introduction The prevention of common diseases relies on identifying The prevention of common and chronic diseases, such risk factors and implementing intervention in high-risk as coronary heart disease, has been complicated by the groups. Nevertheless, most known risk factors have low multifactorial nature of these diseases (Badimon 1993). positive predictive value (PPV) and low population-at- While the classical epidemiological paradigm of search- tributable fraction (PAF) for diseases (e.g., cholesterol ing for "risk factors" and intervening in high-risk groups and coronary heart disease). With advancing genetic has enjoyed much success in controlling and preventing technology, it will be possible to refine the risk-factor many infectious diseases, its success in preventing approach to target intervention to individuals with risk chronic diseases has been mixed. For example, >270 factors who also carry disease-susceptibility allele(s). We factors have been suggested as risk factors for coronary provide an epidemiological approach to assess the im- heart disease (Stehbens 1992). Even to date, there is pact of genetic testing on the PPV and PAF associated continued discussion about the prevention usefulness of with risk factors. Under plausible models of interaction some well-established coronary-heart-disease risk fac- between a risk factor and a genotype, we derive values tors, such as diet and hypercholesterolemia (LaRosa of PPV and PAF associated with the joint effects of a 1992; Allred 1993; Hamsten 1993; Woodard 1993). It risk factor and a genotype. The use of genetic testing has been shown that, for most risk factors for chronic can markedly increase the PPV of a risk factor. PPV diseases, their utility in predicting disease is limited increases with increasing genotype-risk factor interac- (Khoury et al. 1985; Baron 1989). Also, the sensitivity tion and increasing marginal relative risk associated of such factors could be small, limiting both their clinical with the factor, but it is inversely proportional to the relevance and their potential for prevention (Khoury et prevalences of the genotype and the factor. For example, al. 1985). for a disease with lifetime risk of 1%, if all the risk- The new era of molecular genetics and the Human factor effect is confined to individuals with a susceptible Genome Project promises the identification of numerous genotype, a risk factor with 10% prevalence and disease genes and their allelic variants that, per se, may not relative risk of 2 in the population will have a disease cause disease but may interact with other genes and PPV of 1.8%, but it will have a PPV of 91.8% among environmental factors in causing disease (i.e., they may persons with a genotype of 1 % prevalence. On the other be genetic risk factors). There is a growing interest in hand, genetic testing and restriction of preventive mea- the concept of gene-environment interaction in disease sures to those susceptible may decrease the PAF of the causation (Balant et al. 1992; Hegele 1992; Khoury et al. risk factor, especially at low prevalences of the risk fac- 1993). To highlight the importance of gene-environment tor and genotype. With advances in the Human Genome interaction in coronary heart disease, Hegele (1992, p. Project, medicine and public health should consider the 177) states that "some vegetarians with 'acceptable' feasibility of this approach as a new paradigm for dis- cholesterol levels suffer myocardial infarction in the ease prevention. 30's. Other individuals...seem to live forever despite per- sonal stress, smoking, obesity, and poor adherence to a Heart Association-approved diet." To emphasize the Received October 17, 1994; accepted for publication January 5, importance of genetics in nutrition, Simopoulus et al. 1995. (1993, cover) state, in a recently published book on Address for correspondence and reprints: Dr. Muin J. Khoury, Birth genetic nutrition, that "your genes can tell you what to Defects and Genetic Diseases Branch, National Center for Environ- eat and avoid-to live a longer, healthier life." Further- mental Health, Centers for Disease Control and Prevention, F45, 1600 more, it recently has been predicted that "the day of the Clifton Road, Atlanta, GA 30333. © 1995 by The American Society of Human Genetics. All rights reserved. personal DNA profile provided at birth, complete with 0002-9297/95/5604-0004$02.00 calculated risks of various cancers, heart disease, and 835 836 Am. J. Hum. Genet. 56:835-844, 1995 many other conditions could be an actuality by the time Table I that current first-year medical students begin to practice medicine" (Hoffman 1994, p. 130). Parameters of a Genotype-Risk Factor Interaction Model Even if all 50,000-100,000 human genes are cloned of Disease Risk and their functions discovered, and even if sensitive and Susceptibility Risk Relative specific tests are available, it is still far from clear Genotype (g) Factor (e) Prevalence Risk whether we can use genetics to predict disease risk, espe- cially for multifactorial common disorders. One concern Absent ....... Absent (1 - g)(1 - e) 1 Absent ....... Present (1 - g)e Re over the use of genetic testing in disease prediction is Present ....... Absent g(l - e) Rg the notion that genetic tests will have poor disease-pre- Present ....... Present ge Rge dictive ability as well as poor sensitivity (Khoury et al. 1985; Holtzman 1992). Another major concern is re- NoTE.-Exposure and genotype are independent. d = (1 - e) x P(D no factor) + eP(D factor); marginal relative risk associated lated to ethical issues in using genetic testing (Holtzman with exposure is R = P(D factor /P(D no factor); replacing the value 1989; Suzuki and Knudston 1989; Hubbard and Wald of P(D factor) from the second equation into the first equation, we 1993; Garver and Garver 1994). These concerns have obtain P(D no factor) = d/[(l + e(R - 1)]; P(D factor) or PPV (e) led to the development of the Ethical, Legal, and Social = dR/[1 + e(R - 1)]; R = [(1 - g)Re + gRgel[(1 - g) + gRg]. The Issues Program (ELSI) within the National Center for apparent relative risk for genotype-risk factor combination is Rge Human Genome Research (Hoffman 1994). = [(1 - ge)Rge]/[(1 - g)(1 - e) + (1 - g)eRe + g( - e)Rg]. With advancing genetic technology and ongoing soci- etal discussion regarding the use of genetic testing, will it be possible to refine the risk-factor approach to target PAF(e) = e(R - 1)/[1 + e(R -1)] . (2) preventive measures to individuals with risk factors who Throughout our illustrations, we use examples of dis- also carry disease-susceptibility allele(s)? In this article, eases with lifetime risks ranging from 0.1% (such as for we provide an epidemiological assessment of the value some specific birth defects) to 10% (such as for some of using genetic tests to improve the predictive value of specific cancers). For a disease with lifetime risk of 1%, disease risk factors. Using simple epidemiological pa- if a risk factor with 10% prevalence has a relative risk rameters and several plausible schemes of genotype-risk of 2, then we can calculate that PPV(e) = 1.8% and factor interaction, we show the impact of appropriate PAF(e) = 9.1%. Such a risk factor, typical of many genetic testing on the positive predictive value (PPV) and chronic-disease risk factors, is poorly predictive of the the population-attributable fraction (PAF) associated disease and also accounts for a small fraction of cases with disease risk factors. (etiologic heterogeneity). For simplicity, we assume that the risk factor is dichotomous (present/absent) although Methods dose-response functions can be addressed. Marginfal Effects of a Risk Factor Genotype-Risk Factor Interaction Given a disease with a lifetime risk of d in the popula- We assume that, underlying the marginal effects of a tion, suppose that a risk factor (an exposure with preva- risk factor, there exists a pattern of a genotype-risk lence e) is associated with the risk of the disease with factor interaction in producing disease. For simplicity, a relative risk R (equivalent to an odds ratio for rare we assume a measurable genotype (which could be one conditions). For this paper, the clinical and public health allele at one locus or multiple alleles at multiple loci) impacts of the factor on disease occurrence can be mea- with a prevalence g independent of the risk factor. The sured in two ways: joint effects of risk factor and genotype on disease risk are shown in table 1. Three values of relative risk can 1. PPV of the factor-PPV(e): this refers to the risk of be calculated relative to individuals with neither suscep- disease among individuals with the factor (the clini- tibility genotype nor risk factor: cal impact). As shown in table can 1, PPV(e) be writ- Re: relative risk of disease with factor alone (i.e., no ten as susceptibility genotype); *Rg: relative risk of disease with genotype alone (i.e., PPV(e) = d.R/[1 + e(R - 1)] . (1) no risk factor); *Rge: relative risk of disease with both factor and geno- 2. PAF of the factor-PAF(e): this refers to the propor- type. tion of cases that could be prevented if the factor The background risk of disease for people with neither was absent (the public health impact).